Vertical density distributions of fish: a balance between environmental and physiological limitation

Transcription

1 ICES Journal of Marine Science, 59: doi:1.6/jmsc , available online at on Vertical density distributions of fish: a balance between environmental and physiological limitation Boonchai K. Stensholt, Asgeir Aglen, Sigbjørn Mehl, and Eivind Stensholt Stensholt, B. K., Aglen, A., Mehl, S., and Stensholt, E. 2. Vertical density distributions of fish: a balance between environmental and physiological limitation. ICES Journal of Marine Science, 59: Data (trawl, acoustic, CTD) from scientific surveys along the Norwegian coast, in the North Sea, in the Barents Sea, west of the British Isles, and in the Irminger Sea are used. The vertical density distributions of blue whiting, cod, haddock, redfish, saithe, capelin, and herring are described in relation to environmental conditions and physiological limitations. The first four surveys mainly cover banks and shelf areas shallower than 5 m. The last two surveys, aimed at blue whiting and redfish, mainly cover shelf edge and deep-sea areas with depths from to 13 m and from 4 to 3 m. In regard to cod some information from data-storage tags is used. For physoclists the relative vertical profile of each acoustic sample i.e. acoustic-area backscattering coefficient ( ), is expressed in terms of the relative pressure reduction level from seabed up to surface. Thus relative vertical profiles with different bottom depths are normalized and are made compatible for a discussion in terms of the free vertical range (FVR). This restriction to rapid vertical movement is evident in the physoclist species studied. For samples in the shelf area, the profiles show that blue whiting, haddock, saithe, cod, and redfish are mainly distributed within the bottom half of the water column. Some fish adapt to pelagic living especially in areas with high acoustic density and where the bottom is deep. Here a pelagically living fish is defined as an individual fish having a current free vertical range that does not include the seabed. For demersal fish, day and relative vertical profiles are corrected for unequal day and losses in the bottom acoustic dead zone, which is the zone near the seabed where echoes from fish cannot be discriminated from the sea bottom echo. Day and samples are separated by the sun s passing 5 below the horizon. In most years evidence of diurnal vertical migration is found for all investigated species. In many cases of demersal fish there is a higher relative acoustic density ( -values) in the mid-range of the bottom half of the water column in the daytime as opposed to the -time. At there is a degree of separation, one group of fish descends to aggregate near the seabed and another ascends. Inter-annual variations in the diel movement from different parts of the stock are discussed in relation to the inter-annual variations in age composition of the stock. 2 International Council for the Exploration of the Sea. Published by Elsevier Science Ltd. All rights reserved. Keywords: area backscattering coefficient ( ), acoustic bottom dead zone loss, demersal, free vertical range, fish vertical distribution, hydrostatic pressure, pelagic, physostomes, physoclists, swimbladder. Received 29 August 1; accepted 8 April 2. Boonchai K. Stensholt, Asgeir Aglen, and Sigbjørn Mehl: Institute of Marine Research, PO Box 187, N-5817, Bergen, Norway; tel: ; fax: ; boonchai@imr.no. Eivind Stensholt: The Norwegian School of Economics and Business Administration, Helleveien 3, N-545 Bergen, Norway. Introduction In marine teleosts with a swimbladder, that feature occupies % of the body volume (Alexander, 1966; Marshall, 1972). One of its functions is to be a buoyancy organ (Blaxter and Tytler, 1978). However, its physiological limitation in adaptation to pressure changes imposes restrictions on vertical migration (Harden Jones, 1949, 1951, 1952; Brawn, 1962; Kutchai and Steen, 1971; Blaxter and Tytler, 1972, 1978; Ross, 1979; /2/ $35./ 2 International Council for the Exploration of the Sea. Published by Elsevier Science Ltd. All rights reserved.

2 6 B. K. Stensholt et al. Harden Jones and Scholes, 1981, 1985; Arnold and Greer Walker, 1992). There are metabolic costs of swimbladder adaptation and of compensatory swimming when the buoyancy is not neutral (Alexander, 197, 1972; Blaxter and Tytler, 1978). During pressure reduction fish with open swimbladders (physostomes, e.g. herring and capelin), may release gas through a pneumatic duct. Gas secretion is known to be slow and insignificant in physostomes. Brawn s experiment indicates that herring take in air at the surface. The amount of air determines the level of neutral buoyancy, which may be no more than 5 to m. In experiments physostomes respond to rapid pressure increases by ascending (Brawn, 1962; Blaxter and Tytler, 1978; Sundnes and Bratland, 1972). Fish with closed swimbladders (physoclists, e.g. blue whiting, cod, haddock, redfish and saithe) produce gas by secretion in response to under-buoyancy and remove gas by resorption in response to over-buoyancy but these processes take time (Blaxter and Tytler, 1978; Harden Jones and Scholes, 1985). At any point in time the gas content determines a depth level of neutral buoyancy. For rapid vertical movement around that depth is a zone of free vertical range (FVR). Within this range the fish can swim freely and compensate for the deviation from neutral buoyancy by means of swimming movements (Harden Jones, 1949, 1951, 1952; Harden Jones and Scholes, 1981, 1985; Tytler and Blaxter, 1973, 1977; Blaxter and Tytler, 1972, 1978; Arnold and Greer Walker, 1992). By ascending rapidly too far above the neutral buoyancy plane, a physoclist gets too overbuoyant, risking an uncontrollable rise to the surface or a swimbladder rupture (Harden Jones, 1951, 1952; Tytler and Blaxter, 1973; Harden Jones and Scholes, 1985), while physostomes get rid of surplus gas through their duct (Brawn, 1962). On the other hand, during deep dives physoclists have an advantage over physostomes because their gas secretion reduces underbuoyancy. Thus physoclists and physostomes are naturally disposed towards different patterns of vertical distribution and vertical migration. Given the current gas content of the swimbladder, the current free vertical range (FVR) allows rapid vertical movement between two planes, at M and m metres, the limiting depth levels for under- and over-buoyancy. As one atmospheric pressure is about the pressure of a 1 m water column, the pressure ratio (m+1)/(m+1) can be considered a theoretical physiological limit to pressure changes within a free vertical range. Experiments (Harden Jones and Scholes, 1985) have determined that the ratio is ca. 5% in the case of cod. Blaxter and Tytler (1972) concluded that the sensitivity to pressure decrease was about the same for cod and saithe, and that haddock was slightly more sensitive. We expect that saithe, haddock, redfish and blue whiting have a free vertical range which is not very different from that of cod. Several studies conclude that a physoclist, which undertakes extensive diel vertical migrations, generally maintains neutral buoyancy only near the top of its daily vertical range and otherwise is negatively buoyant. Descending some distance below the current free vertical range has no serious consequences however. It may adapt at the bottom by gas secretion and, perhaps with an extra effort, ascend into its FVR (Alexander, 1966; Harden Jones and Scholes, 1981, 1985; Arnold and Greer Walker, 1992; Rose and Porter, 1996; Righton et al., 1). Because of gas loss by passive diffusion due to permeability of the swimbladder wall, there may be a maximal depth where neutral buoyancy maintenance is possible (Kutchai and Steen, 1971; Ross, 1976, 1979; Blaxter and Tytler, 1978). Temperature and salinity impose environmental barriers to fish distribution. In cod a physiological adaptation process starts when it enters waters colder than 2 C that enables it to live in colder temperatures. It also tolerates cold shocks from 4 C to C in normal seawater. High temperatures (above 1 C) with low salinity may be lethal for large cod (Woodhead and Woodhead 1959, 1965; Harden Jones and Scholes, 1974). Björnsson et al. (1) found that large Icelandic cod ( 2 kg) cannot tolerate prolonged exposure to 16 C with salinity 32. The optimal temperature for cod s growth found in laboratory experiments is higher than its natural ambient temperature (Björnsson et al., 1). In the natural environment preferred temperature is just one of many concerns, like availability of food and avoidance of predation (Kristiansen et al., 1). Time series of depth and temperature from data storage tags (DST) attached to adult cod show interplay between the individual cod s migration behaviour and season, depth, temperature, tidal current and its physiological limitation (Stensholt, 1; Righton et al. 1). Combined information from different sources helps the interpretation of conventional fisheries survey data (Righton et al., 1). In the present study information from acoustic records, i.e. the area backscattering coefficient ( )as defined in MacLennan et al. (1), trawl sampling, and CTD, are combined to investigate the distribution of cod, haddock, saithe, redfish, blue whiting, herring, and capelin in relation to environmental and physiological factors. Besides the swimbladder physiology described above, daylight, food availability, predator avoidance, tidal currents, temperature, salinity, depth, and vessel noise are among the factors that influence vertical migration (Neilson and Perry, 199; Ona and Godø, 199; Godø et al., 1999; Aglen, 1994). The observed large-scale vertical distribution is expressed in physiological terms, thus all samples with various depths are normalized, and on that basis the influence of some environmental

3 Vertical density distributions of fish 681 factors is discussed. The structure of the spatial density distribution of fish in relation to environment and physiological limitation can be established. This can be used in the estimation of total stock abundance. Vertical migration behaviour affects the accuracy of acoustic stock estimates of demersal physoclists, mainly through the loss of acoustic fish information in the bottom dead zone, i.e. fish very close to the bottom cannot be distinguished from the bottom itself (Aglen, 1994; Ona and Mitson, 1996; Aglen et al., 1999). Moreover, the extent of losses due to negative buoyancy near the bottom, body orientation, and avoidance reactions to vessel passage among cod, haddock and saithe is still unclear (Ona and Godø, 199; Aglen, 1994; Godø et al., 1999; Totland, pers. comm.). The accuracy can be improved when the amount that is lost varies according to predictable environmental conditions, e.g. daylight and tidal current. How should one interpret vertical profiles, which are based on the observed acoustic abundance of the variable proportion of fish above the dead zone? The difference between the losses in acoustic density for day and is estimated and the vertical profiles are adjusted accordingly, under the assumption that the dead zone is the main reason for day differences in acoustic observations for demersal fish during winter in the Barents Sea (Aglen et al., 1999). In the case of cod the results are discussed in the light of information from data-storage tags, which provides details of the daily vertical ranges and the time spent at each depth level. Materials and methods The Institute of Marine Research, Bergen, performs several annual acoustic surveys for fish stock assessment. In this study observations of acoustic backscattering ( -values) from several acoustic surveys aimed at demersal fish and blue whiting have been used. Data from the demersal fish surveys contain acoustic values that are allocated to the demersal species cod, haddock, saithe and redfish, and, in addition, values allocated to the pelagic species blue whiting, capelin, and herring. In the demersal fish surveys the latter are non-target species and not completely covered. The data considered in this paper are from annual surveys in the main. Firstly those for saithe, haddock, blue whiting, and herring from the saithe surveys along the Norwegian coast from 62 N to 72 N during October November, 1992 (Nedreaas, 1995; Korsbrekke and Mehl, ) and from demersal fish surveys in the North Sea from 55 N to62 N during October November, (Anon., ) and during August September in 1999 (Anon., 1999) (Figure 1). Secondly those for cod, haddock, redfish, capelin, and herring from demersal fish surveys in the Barents Sea, one during February March (Mehl and Nakken, 1996; Mehl, 1997, 1998, 1999; Aglen, et al., 1; Aglen, 1b, 2) (Figure 2) and one during July August 1995 (Aglen and Nakken, 1996; Aglen, 1999, 1a; Michalsen et al., ) (Figure 3). Thirdly from surveys of the spawning stock of blue whiting found west of the British Isles during March April (Monstad et al., 1995, 1996) and finally a survey of oceanic redfish in the Irminger Sea, southeast of Greenland, during June July 1 (Dalen and Nedreaas, 1). The saithe survey is a pure acoustic survey, which means that the decision regarding when and where to make trawl hauls (both demersal and pelagic) is largely based on the acoustic observations. This survey covers areas with highly variable topography, and the survey design therefore varies between regions (Figure 1). The other demersal fish surveys are combined bottom-trawl and acoustic surveys, where there is a fixed grid of predetermined bottom-trawl positions used both for a swept-area estimate and for the interpretation of the acoustic data. Some additional hauls (both demersal and pelagic) are made for identifying acoustic recordings and these are only used for the acoustic estimate. These surveys mainly follow parallel transects with regular spacing. In all the surveys CTD measurements (vertical profiles of salinity and temperature) are made at nearly all trawl positions. The exceptions to this are the Barents Sea summer surveys in 1996 and 1997, and the Barents Sea winter survey in, where parts of the area were covered by hired fishing vessels not equipped for CTD measurements. Typical survey-line transects and trawl stations for each of the demersal fish surveys are shown in bathymetric maps in Figures 1 3. They cover mainly the banks and shelf areas of depth < m. The blue whiting survey is a combined pelagic trawl, bottom trawl, and acoustic survey (Monstad et al., 1995, 1996), which cover mainly the banks and shelf edge of depth from to 1 m. The oceanic redfish (Sebastes mentella) survey is a pelagic trawl and acoustic survey of depths from 5 to 3 m. The -values are split into the redfish within and above the deep scattering layer and, at the same time, above 5 m, and the rest of the redfish because redfish biologists suspect that there are two separate stocks. The -values within and below the deep scattering layer are underestimated by 34% and 55% respectively according to data from a deep towed vehicle (Magnússon et al., 1996; Dalen and Nedreaas, 1). In order to investigate the relationship between fish and environmental variables satisfactorily the respective data must be located at the same locations over the entire studied area. The values at un-sampled locations of temperature, salinity, sea bottom depth, acoustic -values, and density of the studied species, are estimated on the basis of observed values from

4 682 B. K. Stensholt et al Longitude E m 25 3 Norway Norwegian Trench m 1996 Figure acoustic surveys in the North Sea south of 62 N, and along the Norwegian coast north of 62 N. Trawl stations are marked with Δ. Depth isolines are in metres. neighbouring locations by application of geostatistical methods (Cressie, 1991; Stensholt and Sunnanå, 1996). They are available in the software ISATIS (ISATIS, 1997), e.g. kriging with a moving window. Acoustic data and vertical density profile Area backscattering ( -values) were sampled with echointegrators (Knudsen, 199; Foote et al., 1991). During the surveys the acoustic data were scrutinized and the -values were allocated to a number of categories (species or groups of species). For each survey series there is a list of standard categories for the target species and then several non-target species are defined as separate categories usually. Cod, haddock, redfish, herring, and capelin are defined as separate categories in the Barents Sea surveys, while values representing saithe are merged with several non-target species in the category other demersal. Saithe, haddock and blue whiting are defined as separate categories in the coastal saithe survey, the North Sea survey, and the blue whiting survey. During daily scrutinizing onboard the survey vessel, the post-processing system (BEI, Knudsen, 199) presents the echogram data (from surface to bottom) in blocks of 5 nautical miles (nmi) horizontal. Within such a block the user can define a number of sub-blocks (regular or irregular depth intervals or rectangular boxes). Within each sub-block the operator partitions the -values (averaged over 5 nmi) into the separate categories by using the trawl catch species composition of neighbouring stations and subjective judgment based on experience. Thus for each acoustic sample within the same sub-block all species have the same vertical density profile (in relative terms). When the scrutinized data are stored in the database single values are averaged within blocks of 1 nmi (.1 nmi for blue whiting surveys) horizontal resolution and 1 m (25 m for oceanic redfish) vertical resolution, relating to the sea surface. In addition, the values from bottom up to 1 m above bottom are stored with 1 m vertical resolution. These profiles are used for investigating the vertical distributions of fish in relation to environmental factors and physiological limitations. Cumulative relative frequency To investigate the vertical distribution of a physostome species or physoclists with pelagic distribution we selected the acoustic samples with reasonably large total

5 Vertical density distributions of fish 683 Spitsbergen Longitude E Bear Island Svalbard B 3 Great B Norway Novaya Zemlya Russia Summer Figure winter survey in the Barents Sea. Trawl stations are marked with Δ. Depth isolines are in metres. aggregated -values over the water column, as in Stensholt and Nakken (1). To normalize the large spatial variation in absolute values, for each sample s the relative cumulative acoustic abundance C(s,v) of fish is calculated in 1 m depth intervals from the surface down to 1 v metres, v=1, 2,..., D, with bottom depth in the interval (1 (D-1), 1 D]. For each depth step (v=1, 2,...) the calculations of C(s,v) need samples with bottom depth at least 1 v metres, and thus the number of samples is reduced with increasing v. To correct for unequal bottom depth, the water column is partitioned according to fractions of the bottom depth, and the relative -values are accumulated from the surface down to the chosen fractions of the bottom depth. For each level of depth (Figure 16f) or relative depth (Figure 17) the distribution over all samples of the relative cumulative values are presented as box-plots. The boxes indicate the inter-quartile ranges and a line inside each box marks the median. Relative pressure reduction For physoclistous fish the extent of rapid vertical movement is restricted because of their closed swimbladder. At any depth level, the change of swimbladder volume and change of buoyancy status depend on the relative change of pressure. For cod adapted to neutral buoyancy at depth d, the free vertical range (FVR) was determined experimentally as the range between 25% pressure reduction (at depth m) and 5% pressure increase (at depth M) when compared to the pressure at depth d (Harden Jones and Scholes, 1985). Thus the extent of FVR increases with depth. With depths in metres, the pressure ratios are (m+1) (d+1) 1 =.75, (M+1) (d+1) 1 =1.5, and (M+1) (m+1) 1 =2 Thus m+1=.5 (M+1), i.e. the pressure is reduced by 5% for fish ascending from M to m. A full free vertical range for cod with the deeper limit at M () metres stretches up the water column to m=m/2 5 m. In particular, when M is the bottom depth, the bottom restricted free vertical range (FVR bot ) stretches from the seabed up to 5 m above the mid-water level. For other physoclistous species the free vertical range may have a different size, i.e. from m to M where M+1=K (m+1), and where K may be different from 2, but for the species considered here the results below indicate that it is not much different. The free vertical range of physoclistous fish adapted to pelagic living, e.g. schooling fish such as blue whiting in the deep sea, may be calculated using m equal to the

6 684 B. K. Stensholt et al. 5 Spitsbergen Longitude E Great B Norway Novaya Zemlya Russia Summer Figure summer survey in the Barents Sea. Trawl stations are marked with Δ. Depth isolines are in metres. N shows the release site of tagged cod. minimum depth of the distribution layer or M equal to the maximum depth of the distribution layer. The acoustic density coefficient ( -value) depends on the fish buoyancy status. Moreover the acoustic sampling unit represents a rather short time interval, typically about 6 min per nautical mile along the transect line. Therefore, a description of the vertical distribution of acoustic density should be related to the physiological limitation permitting rapid pressure change and, consequently, rapid vertical migration for individual physoclistous fish. In the case of demersal physoclists, it is most useful to investigate the relative vertical profiles of -values in terms of relative pressure reduction with reference to bottom pressure. Here the bottom-restricted, free vertical range (FVR bot ) is defined as the FVR with its deepest end (M) at the seabed, and its extension depends on the fish physiology. It is used as a yardstick for measuring the height above the seabed for investigation of the acoustic vertical profiles. Since m+1=(m+1)k 1,a fish that moves up one FVR experiences a relative pressure reduction of 1 K 1 (K=2 for cod). The height above seabed is partitioned into steps of the relative pressure reduction level (RPRL) with reference to the bottom pressure. The samples with various depth ranges due to unequal bottom depths are transformed to have ranges from RPRL=, at the seabed with depth M, to M (M+1) 1, at the sea surface. In this study the relative pressure reduction at the sea surface can be considered to be approximately 1 in all cases. Consequently, all samples with various depth ranges have the bottom restricted FVR stretching from RPRL= to RPRL=1 K 1 (RPRL=.5 in the case of cod). This normalization allows the comparison of relative profiles from locations with different bottom depths. Then the factors affecting the vertical distribution can be investigated. The graphs show the extent of the distribution range and the vertical distribution profile within each distribution range in terms of an FVR and in relation to the bottom. In the case of a pelagic distribution layer of a physoclist species over great bottom depth, the reference is to the deepest boundary of the distribution range instead of bottom depth. The graphs show the extent of the distribution range and the vertical distribution profile within each distribution range in terms of an FVR, which is independent of the depth level. Relative profiles To normalize for the large variation in acoustic abundance, the profile of each sample is converted into a

7 Vertical density distributions of fish 685 y 75 y 5 Box marks interquartile range, 25 75, of the distribution. Line inside the box marks median y 25 cod. x Seabed RPRL Figure 4. The horizontal axis shows the water column partitioned from seabed to surface according to the relative pressure reduction level (RPRL) with reference to the pressure at the seabed. For each acoustic sample consisting of area backscattering coefficients ( -values), the relative acoustic density is accumulated in a stepwise manner from the seabed up to each RPRL. At RPRL=x, the distribution over the entire sample set of the relative cumulative values up to RPRL=x is presented as a thin box, showing the inter-quartile range and median. The relative cumulative distribution up to RPRL=x has median=y 5, lower quartile=y 25, and upper quartile=y 75, as shown on the vertical axis. Reference line indicates RPRL=.5. relative profile with the observed total -value in the water column as %. Accumulating the relative -value stepwise from the bottom up along the RPRL axis, it grows from to % as RPRL increases from to 1. In most of the cases studied here the fish are demersal, and the increments of the relative cumulative vertical profiles start from the seabed. The range of RPRL, where the relative cumulative value grows from % until it reaches % for most of the samples, may indicate the presence of a physiological restriction in rapid vertical movement due to the relative change of pressure (Figure 4). This range of physiological restriction may correspond to the free vertical range for the species, if no other environmental factors influence the vertical distribution. In the case of pelagic distribution layers, the fish vertical distribution is off the bottom entirely, as for redfish and blue whiting shoals in the deep sea. Thus the relative cumulative vertical profiles should be started from the deepest end of each sample distribution range (Figures 9e,f and 16a e). When the profiles are described in terms of the relative pressure reduction with respect to the pressure at seabed the vertical extent of the school stretches from RPRL=X M upward to RPRL=X m,as in the samples from 1995, 1996, 1998 of Figure 15f. In order to compare the difference X m X M to the FVR in terms of relative pressure reduction, it must be converted and expressed in terms of relative pressure reduction from the deeper to the upper end of the distribution range as a ratio (X m X M ) (1 X M ) 1 = [(1 X M ) (1 X m )] (1 X M ) 1 = 1 [(1 X m ) (1 X M ) 1 ]. The last form corresponds to and may be compared to 1 K 1 above. Adaptation to pelagic life For a physoclist, a rapid ascent too far above the upper limit of its free vertical range is much more hazardous

8 686 B. K. Stensholt et al. than a rapid descent below the lower limit (Tytler and Blaxter, 1973). The latter results in temporary, excessive under-buoyancy (Harden Jones, 1951, 1952; Harden Jones and Scholes, 1981, 1985). When the observed relative cumulative profile reaches % beyond one bottom restricted FVR unit, there are samples with fish detected more than one such unit above the sea bottom. Based on the assumptions that the FVR restriction to rapid vertical movement is effective, and that such a fish has not ascended above its current FVR, we conclude that the fish have adapted to pelagic living in the sense that their current free vertical range does not include the seabed. Fish that are detected within the bottom-restricted FVR may well have an individual, current FVR that does not include the seabed. This means that they just happen to be in the part of their current FVR that partially overlaps with the bottom-restricted FVR, or that they have accepted excessive negative buoyancy. It is impossible to recognize from the acoustic signals if there are fish adapted to pelagic living among those detected within the bottom-restricted FVR. Therefore it is necessary to use the very strict criterion that some fish must have been detected above one bottom-restricted FVR before concluding that part of the stock has adapted to pelagic living. If the interval [m, M] is the bottom-restricted FVR, and there are fish with individual current free vertical range [c, C], where c<m<c<m, then some of them may happen to be detected in the depth interval [m, C], and others, with excessive negative buoyancy, in [C, M]. Only fish that are detected in [c, m] satisfy the criterion. If a fish detected in [m, M] actually has adapted to pelagic living, this cannot be verified from the observed samples. Distribution of relative cumulative -values In Figure 4 the water column is partitioned into RPRL steps of.5 from bottom to surface, shown here along the horizontal axis. The word bottom here means the seabed in case of demersal distribution and the deepest end of each sample distribution range for a clear pelagic distribution. At each discrete RPRL, the box-plot shows the distribution of the observed relative cumulative -values. In each acoustic sample the relative -value is accumulated from bottom up to RPRL=x. The distribution of these relative cumulative values up to each discrete RPRL over the entire sample set is presented as a box showing the inter-quartile range with a line inside each box marking the median. The accumulation levels, in percent, appear on the vertical axis. The distribution of cumulative relative -values up to RPRL=x has median=y 5, lower quartile=y 25, and upper quartile=y 75. For example, the lower quartile=y 25 means that 25% of the samples have the values below, and 75% of the samples have the values above y 25. The line joining the medians for all RPRL is used to represent the relative cumulative acoustic density from bottom up to each RPRL. The inter-quartile range represents the variation of the distribution. If an RPRL has large inter-quartile range, there is large variability among the samples of the relative cumulative acoustic density up to that level. The FVR unit for each species is then estimated from the median, i.e. an RPRL range where the median starts to grow from % until it approaches %, and with the quartiles indicating the uncertainty. As explained above, a sample with fish above the bottom-restricted FVR indicates that some of the fish have adjusted to pelagic living. When the lower quartile is clearly below % at the RPRL corresponding to the upper limit of the bottom-restricted FVR (RPRL=.5 in the case of cod), at least 25% of the samples record fish adapted to pelagic living. Of course this interpretation of the figures builds on the assumption that the defined free vertical range is an effective restriction for rapid vertical movement. The day- time is calculated according to the sun s angle with the horizon, for winter surveys daytime starts and ends when the sun is 5 below horizon. Each acoustic sample is accordingly classified as a day- or -sample, and the vertical profiles may be presented for day and separately and compared. The acoustic bottom dead zone and the day/ cumulative relative frequencies The observed total acoustic -value in the water column may differ significantly between day and. Such differences could be caused by diurnal changes due to fish behaviour and buoyancy status, combined with technical limitations of the hydro-acoustic method to detect fish (Harden Jones and Scholes, 1981, 1985; Aglen, 1994). Thus the basis for calculating relative frequencies changes correspondingly. Based on studies of demersal species during winter in the Barents Sea (Aglen et al., 1999), it is reasonable to assume that the main reason for the difference in day and loss is migration of fish in and out of the near bottom acoustic dead zone. An adjustment for the difference in the day and loss is based on this assumption and that the total amount of fish is equal during day and. Only when the daytime and -time vertical profiles are based on the same proportion of the fish present in the water column, will the comparison of day/ profiles give evidence of fish vertical movements. Thus the observed relative frequencies have been adjusted using the relative change in observed total acoustic -value between day and.

9 Vertical density distributions of fish 687 Starting with one acoustic sample, let A(h) be the observed cumulative acoustic density from seabed up to RPRL=h. With bottom depth b metre, h=1 [(d+1) (b+1) 1 ]=(b d) (b+1) 1 at depth d metres. At the surface, d=, so h=b (b+1) 1, and N=A[b (b+1) 1 ] is the observed total acoustic -value. The observed relative cumulative acoustic density is F(h)=A(h) N 1. Let X be the loss due to the acoustic dead-zone. If the equipment had been able to detect fish at the bottom, A(h) would have been replaced by X+A(h), the total acoustic -value would have been X+N, and the relative cumulative density at level h would have been G(h)=[X+A(h)] (X+N) 1. The relation between the observed F and the correct G is therefore expressed by G(h)=X (X+N) 1 +[N (X+N) 1 ] F(h) =X (X+N) 1 +{1 [X (X+N) 1 ]} F(h) (1) The intercept G()=X (X+N) 1 is the relative loss in the dead-zone, and N (X+N) 1 is the reduction factor for the slope. Increased loss X means higher intercept and lower slope for G. In particular, F()=%, G()=X (X+N) 1, F(b (b+1) 1 )=G(b (b+1) 1 )=%, G (h)=f (h) N (X+N) 1, 1 G(h)=[1 F(h)] N (X+N) 1. For each sample, s, one could draw a similar curve. The observed total acoustic -value is N(s), and the unknown relative loss is X(s) [X(s)+N(s)] 1, which could have been used to correct the curve. All sample curves would have been shifted upward according to Equation (1). This method can be applied to correct for the unequal losses at the bottom due to different reasons, e.g. day-, or tidal current, or before and after the passage of a vessel. We use the median curve as a virtual sample to represent the data. Let virtual dead-zone losses be X d and X n, and observed total acoustic -value be N d and N n for day and, respectively. The total amount of fish at any location does not change between day and, hence X d +N d =X n +N n. For N d and N n respectively, we may use the median of the observed day and total acoustic -value. For the purpose of day and comparison, we only need the difference X n X d =N d N n. Typically the daytime and -time median curves are concave, reach %, cross each other at RPRL=h c (Figure 5), and in an interval around h c the daytime curve is steeper than the -time curve. There is then a higher relative density of fish at day than at in this interval. Before correction, the day and cumulative frequencies are based on different total observed acoustic densities. In order to draw conclusions about vertical migration, one must consider the correction for difference in dead-zone loss, so that both curves are based on the same total acoustic -value. Then one median curve is shifted accordingly. If X n X d =, no correction is needed, and h c remains the same. If X n X d =N d N n >, the correction shifts the -time curve upwards to intercept at (N d N n ) (N d ) 1. This intercept indicates the day difference in the relative acoustic density that is lost due to fish in the dead-zone. With X=N d N n and N=N n, Equation (1) can be rewritten in terms of the observed values as G(h)=(N d N n ) (N d ) 1 + {1 [(N d N n ) (N d ) 1 ]} F(h) =F(h)+[(N d N n ) (N d ) 1 ] [1 F(h)]. Now, in the typical case, daytime accumulation reaches % at a level RPRL=h* where the value is <%. In other words, the median curves of day/ relative cumulative acoustic -values indicate that RPRL=h* is the upper limit of the daytime vertical distribution, but that among samples a significant proportion of fish were still found above RPRL=h*. It is then geometrically clear that the corrected curve still crosses the day curve, say at RPRL=h k, where h c <h k <h* (Figure 5). Hence the median curves show that in fact there must be net evening migrations both upwards and downwards away from the level RPRL=h k. If X n X d =N d N n <, there must be a similar upward shift of the daytime curve (Figure 7). For small shifts the intersections at and h c are replaced by two intersections between and h c, and for large shifts there is no intersection. The normalizations for unequal bottom depths and unequal acoustic densities together with corrections of unequal bottom dead zone loss enable us to compare vertical profiles and discuss vertical migration from the overall line transect survey data. In particular the influence of some environmental factors on the vertical migration can be discussed, e.g. daylight, bottom depth, current and fish density. Results Vertical distribution of fish over different seasons and geographical areas The distribution of relative cumulative acoustic density ( -values) is calculated for all years combined as well as annually for each physoclist and physostome of each

10 688 B. K. Stensholt et al. h k h * day haddock 5 m. h c Figure 5. The median curves of day/ relative cumulative acoustic profiles indicate RPRL=h* as the upper limit of the daytime vertical distribution, but among time samples a significant proportion of fish are still found above RPRL=h*. The adjusted curve (broken line) crosses the day curve at RPRL=h k where h c <h k <h*. The higher slope of the daytime curve around RPRL=h k shows higher density at daytime than -time. Hence the median curves show that there must be net evening migration both upwards and downwards away from the level RPRL=h k. survey. The results are shown in Figures For all physoclists the relative cumulative vertical profiles are started from the seabed, except for redfish when it is pelagic (Figure 9e,f) and for blue whiting in the deep sea (Figure 16a e). In these cases the profiles are started from the deepest boundary of each sample distribution range. The figures are presented for each survey and species, and within each species the acoustic samples are grouped according to day- or to density level and bottom depth range, as indicated at the bottom right corner of the figures. The species are cod and haddock from winter surveys in the Barents Sea (Figures 6 8); redfish from winter surveys in the Barents Sea and Irminger Sea (Figure 9); cod, haddock, and redfish from summer surveys in the Barents Sea (Figures 1 12); saithe, haddock, and blue whiting from surveys along the Norwegian coast (Figures 13, 14); saithe and blue whiting from surveys in the North Sea (Figure 15); blue whiting from surveys west of the British Isles (Figure 16) and capelin and herring from winter surveys in the Barents Sea (Figure 17). The results for demersal fish surveys reflect the vertical distribution of fish found in the shelf sea (Figures 1 3). There are data sets where a few samples have very high density and account for a high proportion of the sampled fish. They have special profiles and this is not reflected in the median of the entire collection of samples. Therefore, such samples are investigated separately. The annual median curves of the relative cumulative -values show the annual variation in vertical distribution (e.g. Figure 6b). The reference line at RPRL=.5 indicates one bottom restricted FVR for cod. For other species, the RPRL where the median gets close to % and the inter-quartile range approaches may indicate the extent of their free vertical ranges. They seem to be of roughly the same size as for cod (Figures 6a, 8a 13a,c,e, 9f, 15a,c, 16d,e,f). Below the reference line

11 Vertical density distributions of fish 689 (a) (b) cod cod (c) (d) day s A cod cod 5 m (e) (f) day cod 5 m cod m Figure 6. Distribution of cod relative cumulative vertical profiles from the seabed to the surface in terms of RPRL, from winter surveys in the Barents Sea, Median and inter-quartile range (represented by thin boxes, left box for day, right box for ) of the distribution, (a) all data; (b) annual median curves; (c) day- and samples; (d) day median curves with upward adjusted median curve (broken line); (e) samples with bottom depth < m; (f) samples with bottom depth m, more cod adapt to pelagic living.

12 69 B. K. Stensholt et al. day cod m Figure 7. Day and median curves of the distributions of cod vertical profiles of relative cumulative acoustic -values in relation to relative pressure reduction level, RPRL, with upward adjusted day median curve (broken line). Samples are from the winter survey in the Barents Sea, Reference line indicates RPRL=.5. and close to the bottom the inter-quartile range is relatively large, indicating a large variation of the relative cumulative values among different acoustic samples. This means that within one FVR the fish are free to distribute along the water column in a way suited to local conditions. Pelagic living for some fish in the samples is indicated when the relative cumulative -values show there are fish higher than RPRL=.5, i.e. the lower quartile reaches % beyond the reference line (Figures 6c,f, 8c, 9c, 1e, 11e, 13c, 15c,e). In such cases more than 25% of the samples have some fish adapted to pelagic living. This is observed especially in samples from areas of high acoustic abundance and where the seabed is deeper than m (Figures 6f, 8c, 9c, 1e, 11e, 14b,d). The most notable species may be oceanic redfish and blue whiting (Figures 15e,f, 16e,f), which are pelagic at great depth. For surveys in the Barents Sea, the distribution at each RPRL has larger inter-quartile range in summer than winter (Figures 6a, 8a 12a). The distribution in summer 1999 shows more pelagic living for cod and haddock than in most years. The depth distribution (a) (b) haddock haddock (c) day (d) haddock haddock 5 m Figure 8. Distribution of haddock relative cumulative vertical profiles from the seabed to the surface in terms of RPRL, from winter surveys in the Barents Sea, Median and inter-quartile range (represented by thin boxes, left box for day, right box for ) of the distribution (a) all data; (b) annual median curves; (c) day- and samples; (d) day median curves with upward adjusted median curve (broken line).

13 Vertical density distributions of fish 691 (a) (b) redfish redfish (c) day (d) redfish redfish 5 m (e) day (f) oceanic redfish 45 3 m oceanic redfish 45 3 m Figure 9. Distribution of redfish relative cumulative vertical profiles from the seabed to the surface in terms of RPRL, from winter surveys in the Barents Sea, Median and inter-quartile range (represented by thin boxes, left box for day, right box for ) of the distribution (a) all data; (b) annual median curves; (c) day- and samples; (d) day median curves with upward adjusted median curve (broken line); relative cumulative profiles from the deepest of the distribution layer to surface of samples from Irminger Sea, 1 (e) entire redfish distribution range; (f) only above and within deep scattering layer.

14 692 B. K. Stensholt et al. (a) (b) cod cod (c) (d) cod cod 5 m 5 m (e) (f) cod m cod m Figure 1. Distribution of cod relative cumulative vertical profiles from the seabed to the surface in terms of RPRL, from summer surveys in the Barents Sea, Median and inter-quartile range (represented by thin boxes) of the distribution (a) all data; (b) annual median curves; (c) and (d) samples with bottom depth< m; (e) and (f) samples with bottom depth m, showing more cod adapt to pelagic living. of individual cod from data storage tags has larger variation in summer than in winter (Tables 1, 2), in agreement with the above results. Proportion of fish in the bottom channel For the physoclist species, the ratio of the observed -value from the bottom channel ( 1m above seabed)

15 Vertical density distributions of fish 693 (a) (b) haddock haddock (c) (d) haddock 5 m haddock 5 m haddock m (e) haddock (f) m Figure 11. Distribution of haddock relative cumulative vertical profiles from the seabed to the surface in terms of RPRL, from summer surveys in the Barents Sea, Median and inter-quartile range (represented by thin boxes) of the distribution (a) all data; (b) annual median curves; (c) and (d) samples with bottom depth< m; (e) and (f) samples with bottom depth m, showing more haddock adapt to pelagic living. to the observed total -value was calculated in each sample and the distribution of the ratio investigated. The median of the distribution is shown in Tables 3 4.A general feature is that in deeper waters fish distributions spread relatively higher up into the water column. Table 3 shows that cod and haddock have a higher proportion in the bottom channel during than day but this is based on the observed acoustic values without adjustment for unequal losses during day and.

16 694 B. K. Stensholt et al. (a) (b) redfish redfish (c) (d) redfish 5 m redfish 5 m (e) (f) redfish m redfish m Figure 12. Distribution of redfish relative cumulative vertical profiles from the seabed to the surface in terms of RPRL, from summer surveys in the Barents Sea, Median and inter-quartile range (represented by thin boxes) of the distribution (a) all data; (b) annual median curves; (c) and (d) samples with bottom depth < m; (e) and (f) samples with bottom depth m. Table 2, based on depth records from data storage tags, shows that individual cod spends more time in the lower part of the daily depth range (within 1 m above the daily maximum depth) during December January than during other periods. When cod is in the lower part it is likely that for some of the time it would have been detected by acoustics and would have contributed to the -value in the bottom channel, and at other times it is in the acoustic bottom dead zone.

17 Vertical density distributions of fish 695 (a) (b) saithe 3 m saithe 3 m (c) (d) day haddock 3 m haddock 3 m (e) day (f) blue whiting 3 m blue whiting 3 m Figure 13. The distribution of relative cumulative vertical profiles from the seabed to the surface in terms of RPRL, from surveys along the Norwegian coast Day and medians and inter-quartile ranges (represented by thin boxes, left box for day, right box for ), from combined data for (a) saithe; (c) haddock; (e) blue whiting. Day median curves with upward adjusted median curve (broken line) for (b) saithe; (d) haddock; (f) blue whiting.

18 696 B. K. Stensholt et al. (a) (b) saithe 3 m >5 saithe 3 m (c) (d) haddock 3 m haddock 3 m (e) (f) blue whiting 3 m blue whiting 3 m Figure 14. The distribution of relative cumulative vertical profiles from the seabed to the surface in terms of RPRL, from surveys along the Norwegian coast Annual median curves for samples with bottom depth < m of (a) saithe; (c) haddock; (e) blue whiting; and for samples with bottom depth 3 m of (b) saithe; (d) haddock; (f) blue whiting. The level of -value indicated at bottom right corner.

19 Vertical density distributions of fish 697 (a) (b) saithe 3 m saithe 3 m (c) (d) blue whiting 3 m blue whiting 3 m (e) (f) blue whiting 3 m blue whiting 3 m Figure 15. The distribution of relative cumulative vertical profiles from the seabed to the surface in terms of RPRL, from surveys in the North Sea Median and inter-quartile range (represented by thin boxes) of (a) saithe; (c) and (e) blue whiting. Annual median curves of (b) saithe; (d) and (f) blue whiting. Acoustic -values level, species and bottom depth ranges are shown at the bottom right corner. 99

20 698 B. K. Stensholt et al. (a) (b) m m (c) (d) m (e) m (f) m m Figure 16. Median (full or broken line join day or median) and inter-quartile range (represented by thin boxes, left box for day, right box for ) of blue whiting from the distribution of relative cumulative vertical profiles from the deepest end of each sample distribution layer to surface in terms of RPRL. Samples are from surveys west of British Isles. Bottom depth and year are as specified at bottom right corner. The pelagic distribution layers are shown in (f), with accumulation from surface down in metres.

21 Vertical density distributions of fish 699 (a) (d) capelin herring (b) (e) capelin herring (c) (f) day capelin herring Figure 17. The distribution of relative cumulative vertical profiles from the surface to the bottom in terms of relative depth (. at surface, 1. at bottom), from winter surveys in the Barents Sea, Median and inter-quartile range (represented by thin boxes, left box for day, right box for ) of the distribution for (a) capelin; (d) herring; annual median curves (b) capelin; (e) herring; day and for (c) capelin; (f) herring. Dead-zone loss during day and The typical case in the Barents Sea winter surveys was that the observed acoustic abundance during the daytime was higher than at, N d >N n, i.e. that -time loss in the acoustic dead-zone exceeds daytime loss. The -time median curve is adjusted by an upward shift. The values for N d and N n are the medians of the total acoustic -value in the water column over the day and samples, respectively. Only the samples with total -value greater than 1 are used here.

22 7 B. K. Stensholt et al. Table 1. Temperature and depth distribution, in winter (January March) and summer (July August). From time series of 6 data-storage tags attached to North East Arctic cod released in the Barents Sea. Tag no. Mean Std Min Max P5 Q1 Median Q3 P95 Temperature 117 Summer Winter Summer Winter Summer Winter Summer Winter Summer Winter Summer Winter Depth 117 Summer Winter Summer Winter Summer Winter Summer Winter Summer Winter Summer Winter For the six winter surveys and the three species cod, haddock, and redfish, the dead zone corrections, i.e. the relative difference in losses at the bottom (N d N n ) (N d ) 1 if N d >N n ( correction), or (N n N d ) (N n ) 1 if N d <N n (day correction), are given in Table 5. The material may be insufficient to allow the conclusion that 1996, with its large day corrections, is an exceptional year (Figure 7). One possibility is that there is a separation of both the cod and the haddock stocks into groups with very different diurnal vertical migration patterns and that switches between N d <N n and N d >N n from one year to another is explained by relative changes in the abundance of such groups. The relative Table 2. Percentage of time spent within 1 m above the daily maximal depth from depth records during April June (4 6), July August (7 8), September November (9 11), December January (12 1), and February March (2 3). Depth time series from April 1996 to March 1997 are from 6 data-storage tags attached to North East Arctic cod released in the Barents Sea. Tag number cumulative acoustic frequency figures for day and (Figures 6d, 8d, 9d) indicate that such separation exists, although the acoustic data give no information on which parts of the stock ascend or descend. The pattern of vertical migration in 1996 corresponds to the high proportion of the one-year-old fish (Table 6), in agreement with Aglen et al. (1999). Unless the species and age allocation of acoustic samples can be made according to age and species composition in smaller depth intervals, such profiles may also occur in areas where different species and age groups have opposite vertical migration. Diel pattern in vertical distribution In most cases, when the samples from winter surveys (cod, haddock, and redfish ) are classified as day or, the relative cumulative distribution of -values near the sea bottom shows lower median values at daytime than at -time. With accumulation further up the water column, the day median increases faster than the median, and the order is reversed around mid-range in the lower half of the water column (Figures 6d, 7d, 8d). This reversal appears as the day median curve s crossing and passing over the median curve, with a steeper slope of the day median curve than the median curve. When the crossing remains after the correction for unequal dead zone loss, it shows that during -time a higher proportion of